The Key Technologies Research Of Modern Signal Processing And Patten Recognition For Mechanical Failture Diagnosis

The process of machinery of fault diagnosis includes the acquisition of information and extracting feature and recognizing condition of which feature extraction and condition identification are the priority. To solve the fault feature extraction problem completely, theories, methods and techniques for information processing, especially modern signal processing, have been depended on, and new methods, new theories and new techniques for fault feature extraction are being researched.Based on these conditions, this paper describes the exploratory work of the fault feature extraction in mechanical systems with the wavelet analysis, lifting wavelet, empirical mode decomposition(EMD), independent component analysis(ICA). The main contents of this paper include:(1) The theory of the lifting wavelet and grey system are applied in the research of the fault diagnosis for machines, based on the lifting wavelet transform for vibration signals, as well as the time and frequency hybrid features pattern is built, and the use of grey incidence degrees as pattern recognition method realized to revolved the machinery precise diagnosis. The experimental results imply that the method is not only effective, but also of great potential in fault diagnosis of machines.(2) This article proposed one new fault diagnosis method of roller bearing: grey incidence analysis of EMD-AR(empirical mode decomposition-autoregresssive) model. First take EMD to the vibration signal to decompose, then recombines the basic pattern component (Intrinsic Mode Function, IMF) establishes the AR model according to the reorganization component, the model from the return parameter and the model residual error variance composition characteristic vector, the use of grey incidence degrees as pattern recognition method realized to revolved the roller bearing precise diagnosis.(3) Based on the analysis of the mathematic model of machine signal, independent component analysis (ICA) technique is utilized to extract the independent components defined as vibratory waveform basis from machine signal. Through projecting the vibratory signal profiles on the wave with big kurtosis value. some independent features are acquired. Because the vibratory waveform basis are actually the vibratory source center responses, extracted features are not only independent each other, but also have real physical meaning. Further, simulation test and comparison test are carried out using support vector machine (SVM), and simulation results demonstrated the validity and feasibility of the proposed method.(4) In order to identify sources from Rotor-Bearing system, analyze the mechanism of nonlinear dynamics for fault Rotor, a method for sources recognition based on wavelet -independent component analysis (ICA) is proposed. By combining wavelet packet and ICA, their advantages are brought into play well, and every independent sources embedded into multi-channel observation by sensors are separated. Experimental results show that this new approach is satisfactory and proposed one new method for recognition of nonlinear vibration response of Rotor-Bearing system.(5) The mechanism of vibration source convolution is analyzed, and a method for bispectra-blind source separation for convolved signals is proposed.(6) Improved PCA for convolved signals separation is proposed. The essential objective of an engineering diagnostic system is to detect the potential faults existing in a continuously running machine.